Transparent methodology
How CryptoVideos turns creator coverage into research evidence
Every published claim must be traceable to a public video timestamp, classified by evidence type and checked against independent coverage or market data.
Automatic creator admission
A candidate channel can earn up to 100 points: 20 for a secure public feed, 20 for a confirmed YouTube channel identity, 10 for enough recent items, 40 for consistent cryptocurrency relevance and 10 for a publishing history that spans at least two weeks. The default admission threshold is 70. Passing it permits analysis; it does not make every creator claim trustworthy.
Timestamped claim extraction
The video model must return structured JSON that separates factual statements, opinions and measurable predictions. Every claim needs a stable canonical key, a timestamp and an extraction-confidence score. Missing fields, generic summaries and malformed output fail validation automatically.
Independent agreement
Support and disagreement are counted across distinct creator sources. Repeated uploads from one channel increase the video count but never the creator count. The displayed agreement percentage is the larger of the supporting or disputing share; a tied or unclear result remains unverified.
Contract and market context
When a complete contract address appears in a video, GoPlus risk indicators and Etherscan source-verification data can be attached to the evidence page. CoinGecko provides cached market context, with Coinranking available as a fallback. These signals add context; they never convert a creator’s claim into a fact.
Crypto Conversation Index
The first-party index is calculated from validated video mentions, distinct creator coverage and relevant headlines from connected evidence feeds. The initial weights are three points per video, five per independent creator and 1.5 per news item. Current and preceding windows are compared to show coverage momentum—not market sentiment.
Publication threshold
Invalid structured output, missing timestamps, irrelevant topics, duplicated URLs, provider exhaustion and uncheckable high-risk claims are rejected or held for another attempt. There is deliberately no “publish the source unchanged” fallback.
Why can a published claim still be wrong?
A published item establishes what a creator said and where they said it. Independent agreement helps measure a narrative, but primary documents, on-chain records and observable outcomes remain the stronger evidence.